Lecture 3: Carrier Ethernet Technology · Ethernet priority 802.1p is amendment in 802.1q Allow...
Transcript of Lecture 3: Carrier Ethernet Technology · Ethernet priority 802.1p is amendment in 802.1q Allow...
Lic.(Tech.) Marko Luoma (1/49)
S-38.3191 Verkkopalvelujen tuotanto
S-38.3191 Network Service Provisioning
Lecture 3: Carrier Ethernet Technology
Lic.(Tech.) Marko Luoma (2/49)
Ethernet Two frame types:
Ethernet-II: Address fiels are followed by lenght field
IEEE-802.3: Address fields are followed by type field
Multiprotocol capability» VLAN is a ’protocol’
Frame sizes 64 <-> 1518 Bytes Historically due to collision detection in large flat networks Additionally
IFG: 12bytes Preamble: bytes Start delimeter: 1 byte
DataStart(1 byte)
DST Add(6 bytes)
PADPreamble(7 bytes)
FCS(4 bytes)
SRC Add(6 bytes)
Length(2 bytes)
Lic.(Tech.) Marko Luoma (3/49)
Virtualisation of Ethernet Virtual LAN is a network within the network
It logically creates virtual networks on top of physical network
Virtuality is realized with additional fields in the frame
Tag Protocol ID: 0x8100 for 802.1q
Priority: 802.1p priority
CFI: Canonical format indicator (MAC address can be or not)
VLAN ID: 4096 VLAN IDs
» Only few (read expensive) devices support simultaneously this number
DataStart(1 byte)
DST Add(6 bytes)
PADPreamble(7 bytes)
FCS(4 bytes)
SRC Add(6 bytes)
Length(2 bytes)
Priority(3 bits)
CFI(1 bit)
(0x8100)(2 bytes)
VLAN ID(12 bits)
Lic.(Tech.) Marko Luoma (4/49)
VLAN Separation of network resources to logical units is based on forwarding
databases (FDB)
In independent mode, each VLAN has its own FDB
Clients residing in different forwarding table are not able to communicate without external help
In shared mode, part of VLANs share a common FDB
Clients residing (symmetrically) in same FDB are able to communicate together
Communication between VLANs is established with
‘Misconfigured’ bridge that connects VLANs together
Router forwarding packets between VLANs
Lic.(Tech.) Marko Luoma (5/49)
VLANs PVID ~ Port VLAN identifier
Each and every switchport is assigned to belong to particular VLAN
Incoming untagged traffic is forwarded by using this VLANs FDB
Address learning is bound to that FDB
Incoming tagged traffic is associated to VLAN based on VID or FDB depending on ingress filtering rules
VID ~ VLAN identifier
If frame is
Coming in from a trunk interface it contains 802.1q tag which carries VID
Going out to trunk link packet is coded to 802.1q tag mode
VID usually is PVID from the ingress port
Lic.(Tech.) Marko Luoma (6/49)
Port based VLANs Ingress filtering rules:
Received frame is untagged
Forward using PVID
Discard
Received frame is tagged
Forward using VID
VID = 0:
» Use only P-bits, forward using PVID
VID = 1
» Default tree, all interfaces
Forward using PVID
Discard
Lic.(Tech.) Marko Luoma (7/49)
Port based VLANs Egress filtering rules
Interface is in untagged mode
Forward untagged frame
Use configured priorities
Interface is in tagged mode
Set tag based on classification rules
Ingress VID
PVID
P-bits
Lic.(Tech.) Marko Luoma (8/49)
Ethernet priority 802.1p is amendment in 802.1q
Allow traffic prioritization within Ethernet networks
3 bits -> 8 priorities
Number of queues dependent of HW
At minimum strict priority queuing between queues
Scheduling algorithms are studied in detail later on this course
Mapping traffic to queues is dependent on
Number of queues
Configured policy (egress filtering) MAC address Ethertype DSCP Address
Lic.(Tech.) Marko Luoma (9/49)
Extending Ethernet Ethernet networks can be extended from LANs to MANs by adding new
mechanisms Larger scale virtualization
More separate virtual networks (4096 ???) Separate forwarding index
Hiding LANs from MANs
Lic.(Tech.) Marko Luoma (10/49)
Provider networks Provider networks are extensions to customer networks
Provider Bridged Networks extend customer networks with a additional VLANs Provider bridges
Provider Backbone Bridged Networks extend customer networks with additional address structure Provider backbone bridges
Provider Edge Bridge (PEB)
Provider Bridge (PB)
Backbone Edge Bridge (BEB)
Backbone Core Bridge (BCB)
Customer Equipment (CE)
Provider Bridged Network Provider Backbone Bridged Network
Lic.(Tech.) Marko Luoma (11/49)
Provider network technologies 802.1ad
Q-in-Q
Provider tagging
Enterprise addresses are carried by provider switches service VLANs
Enterprise LANs are aggregated into single provider LAN
802.1ah
M-in-M
Provider encapsulation
Enterprise addresses are invisible for provider switches
Enterprise addresses are aggregated into few provider MACs
ProviderTag#1
ProviderTag#N
EnterpriseHeader
EnterpriseTag
DataProviderQ-Tag
ProviderS-Tag
ProviderHeader
EnterpriseHeader
EnterpriseTag
Data
Lic.(Tech.) Marko Luoma (12/49)
Q-in-Q Frame Format Provider tagging cascades several Q-tags into the frame
Ethertype 0x88a8
Priority is provider dependent not copied from customer settings
Provider Tag = S-Tag = Service Tag
P-VLAN ID = EVC ID
DataC-DST(6 bytes)
(0x88a8)(2 bytes)
C-VLAN ID(12 bits)
C-DST(6 bytes)
FCS(4 bytes)
P-Priority(3 bits)
C-CFI(1 bit)
P-VLAN ID(12 bits)
(0x8100)(2 bytes)
P-CFI(1 bit)
C-Priority (3 bits)
Lic.(Tech.) Marko Luoma (13/49)
M-in-M Frame Format Provider encapsulation allows second layer of operation
With or without Q-in-Q-tag
With or without Q-tag
First tag in M-in-M header is traffic engineering tunnel tag
Ethernet traffic engineering
Second tag in service tag = I-SID = EVC ID
DataC-VLAN ID(12 bits)
C-DST(6 bytes)
C-DST(6 bytes)
C-FCS(4 bytes)
C-CFI(1 bit)
(0x8100)(2 bytes)
C-Priority (3 bits)
P-FCS(4 bytes)
P-DST(6 bytes)
(0x88a8)(2 bytes)
P-DST(6 bytes)
PQ-Priority(3 bits)
PQ-VLAN ID(12 bits)
PQ-CFI(1 bit)
(0x88e7)(2 bytes)
Reserved(7 bits)
Service ID(24 bits)
PT(1 bit)
Lic.(Tech.) Marko Luoma (14/49)
M-in-M Frame Format M-in-M + Q-in-Q allows scalable provisioning of core + metro services
Q-in-Q: Small Metro encapsulation
M-in-M: Large Metro and Core aggregation
Traffic Engineering
State space reduction
DataC-DST(6 bytes)
(0x88a8)(2 bytes)
C-VLAN ID(12 bits)
C-DST(6 bytes)
C-FCS(4 bytes)
P-Priority(3 bits)
C-CFI(1 bit)
P-VLAN ID(12 bits)
(0x8100)(2 bytes)
P-CFI(1 bit)
C-Priority (3 bits)
P-FCS(4 bytes)
P-DST(6 bytes)
(0x88a8)(2 bytes)
P-DST(6 bytes)
PQ-Priority(3 bits)
PQ-VLAN ID(12 bits)
PQ-CFI(1 bit)
(0x88e7)(2 bytes)
Reserved(7 bits)
Service ID(24 bits)
PT(1 bit)
Lic.(Tech.) Marko Luoma (15/49)
Forwarding options What we use as a index in forwarding database
Vanilla case: Destination Address or VID+DA Why to restrict oneself to this
This is just IEEE requirement that one must support legacy Ethernet There will be time, when we must let us free from the burdens of
history Fields that could be used
SA DA VIDs
Lic.(Tech.) Marko Luoma (16/49)
Forwarding options
MAC based Destination Source and Destination
VLAN based Single Cascaded
MAC and VLAN based Source and VLAN Source + Destination + VLAN Source + Cascaded VLAN Source + Destination + Cascaded VLAN
Lic.(Tech.) Marko Luoma (17/49)
MAC based forwarding Single FDB where lookup is based on the MAC addresses
Destination MAC Downstream merging
Source and Destination MAC Independent paths
Lic.(Tech.) Marko Luoma (18/49)
VLAN based forwarding Single FDB where lookup is based on VID values (labels)
Single VID Global values: lookup Link local values: swap
Multiple VID Outer label: forwarding label
Global Link local
Inner label: service label (only visible at the receiver) Multiple FDBs where lookup is based on VID values (labels)
Single VID Partial label space Multicast
Multiple VID Outer label: FDB index (service label) Inner label: Forwarding label
Lic.(Tech.) Marko Luoma (19/49)
MAC and VLAN based forwarding
Multiple FDBs where lookup is based on MAC addresses VID indexes the FDB
Destination MAC lookup Downstream merging
Source + Destination MAC lookup Independent paths
Single FDB where lookup is based on combination of MAC and VLAN Destination + VLAN lookup
Service dependent downstream merging Source + Destination + VLAN lookup
Independent forwarding paths
Lic.(Tech.) Marko Luoma (20/49)
Carrier Grade System/hardware/software that is extremely reliable so that it could be used
as a part of a network that is engineered to deliver high availability services Five nines or more
Contains mechanisms to monitor its internal state and to recover from internal error
Contains mechanisms to cope with external influences Fast recovery (conventionally less than 50ms)
Lic.(Tech.) Marko Luoma (21/49)
Carrier Ethernet Simple idea:
Sell Ethernet services from the carrier infrastructure
Network TechnologySDH/MPLS/IPAccess Technology
Ethernet
Lic.(Tech.) Marko Luoma (22/49)
Carrier Ethernet Transport Idea:
If access is Ethernet, why not make the core network also with Ethernet
Network TechnologyEthernetAccess Technology
Ethernet
Lic.(Tech.) Marko Luoma (23/49)
Ethernet vs Carrier Grade Ethernet
Ethernet FDBs are populated by MACs
learning Topology information is flat Loops are solved by RSTP No internal state monitoring Only one virtualization layer (VID)
CGE FDBs are populated statically by
management Topology information is separated
by access and core Loops are handled by routing Internal state is monitored by OAM Multiple layers of virtualization
QinQ MinM Combinations
Lic.(Tech.) Marko Luoma (24/49)
MetroEthernetForum: Carrier Grade Ethernet
Technology perspective Customer side: semi-transparent Ethernet Provider side: SDH, VPLS, Q-in-Q etc
Service definitions Ethernet line (E-LINE)
point-to-point Ethernet LAN (E-LAN)
multipoint- to-multipoint Ethernet Tree (E-Tree)
point-to-multipoint
Lic.(Tech.) Marko Luoma (25/49)
Major changes Service concept
E-LAN, E-Tree and E-Line Connection orientation
Ethernet Virtual Connection (EVC) Filter word for distinguishing packets from different connections
QoS SLA is required for large scale deployment within corporate
interconnections Bandwidth control
Committed information rate control (inherited form FrameRelay)
Lic.(Tech.) Marko Luoma (26/49)
E-Line and E-LAN Services
E-Line Service used to create
Private Line Services
Ethernet Internet Access
Point-to-Point VPNs
E-LAN Service used to create
Multipoint VPNs
Transparent LAN Service
CE
CE
Point-to-Point EVC
MENUNI
UNI
E-Line Service type
CE
CE
CE
MEN
CE
Multipoint-to-Multipoint EVC
UNI
UNI
UNI
UNI
E-LAN Service typeSource: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (27/49)
Yes or No. Specifies whether customer VLAN ID is preserved or not.CE-VLAN ID
Preservation
Yes or No. Specifies whether customer VLAN CoS (802.1p) is preserved
or not.
CE-VLAN CoS
Preservation
Service Attribute ParametersService Attribute
Specifies the Frame Delay, Frame Jitter and Frame Loss per EVC or
frames within an EVC identified via their CE-VLAN CoS (802.1p) valueService Performance
Discard or Tunnel per protocolLayer 2 Control
Protocol Processing
Specifies whether broadcast frames are Discarded, Delivered
Unconditionally or Delivered Conditionally
Broadcast Service
Frame Delivery
Specifies whether multicast frames are Discarded, Delivered
Unconditionally or Delivered Conditionally
Multicast Service
Frame Delivery
Specifies whether unicast frames are Discarded, Delivered
Unconditionally or Delivered Conditionally
Unicast Service
Frame Delivery
A list of UNIs (identified via the UNI Identifier service attribute) used with
the EVCUNI List
Point-to-Point or Multipoint-to-MultipointEVC Type
Source: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (28/49)
None or <CIR, CBS, EIR, EBS>. This Bandwidth profile applies to all frames marked with a particular CoS ID over an EVC.
Ingress Bandwidth Profile Per CoS ID
The maximum number of EVCs allowed per UNIMax. Number of EVCs
Mapping table of customer VLAN IDs to EVCCE-VLAN ID / EVC Map
Yes or No. Defines whether multiple services can be on the UNIService Multiplexing
A string used to identity of a UNI, e.g., NYCBldg12Rm102Slot22Port3UNI Identifier
Discard, Peer or Pass to EVC per protocolLayer 2 Control Protocol Processing
None or <CIR, CBS, EIR, EBS>. This Bandwidth profile applies to all frames over particular EVC.
Ingress Bandwidth Profile Per EVC
Service Attribute ParametersService Attribute
None or <CIR, CBS, EIR, EBS>. This Bandwidth profile applies to all frames across the UNI.
Ingress Bandwidth Profile Per Ingress UNI
No or Yes (all customer VLAN IDs are mapped to an EVC at the UNI).All to One Bundling
No or Yes. Specifies that one or more customer VLAN IDs are mapped to an EVC at the UNI
Bundling
A string used identify an EVC, e.g., NYCBldg1Rm102Slot22Port3EVC3UNI EVC ID
IEEE 802.3-2002MAC Layer
Full Duplex or Auto negotiationMode
10 Mbps, 100 Mbps, 1 Gbps or 10 GbpsSpeed
Standard Ethernet PHYPhysical Medium
Source: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (29/49)
Scalability• Services and Bandwidth• 100,000’sofEVC’s• From Mbps to x10Gbps
Protection• 50ms Protection• End to End Path Protection• Aggregated Line & Node Protection
Hard QoS• Guaranteed end to end SLA• End to End CIR and EIR• Business, Mobile, Residential
TDM Support• Seamless integration of TDM• Circuit Emulation Services• Support existing voice applications
ServiceManagement• Fast service creation• Carrier class OAM capabilities• Customer Network Management (CNM)
CarrierEthernet
Source: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (30/49)
MEF Vision for NGN
BusinessSubscriber
Service Provider
Transport
IP
IP
• Carries all applications• Internet Access• IP VPN service
• Hard QoS and Traffic Engineering• Better bandwidth utilization/Scalability• Protection Switching (50ms)• OAM functions and TDM Support• Any-to-Any VPN services• Multicast support with Hard QoS
ResidentialSubscriber
Source: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (31/49)
VPNs in CGE Any-to-any Layer 2 VPN to interconnect multiple locations transparently and
effectively (with high and guaranteed bandwidth)
Branch
ISP
IP
Branch
Branch Branch
HQ
Internet
E-LAN
Source: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (32/49)
CGE and IP
IP
IP
E-LAN connections among routers Any logical IP topology, multiple EVC’s on a single physical interface
R R
R
R
R
R
R
R
R
R
R
R
IP NetworkUsing Carrier Ethernetas the Transport Layer
Source: Metro Ethernet Forum
Lic.(Tech.) Marko Luoma (33/49)
CGE and IP IP networks are collapsed in to two layers
Residential customer concentration routers Integrated in DSLAM
Border routers For address propagation and policy control
IP aggregation is vanishing Ethernet aggregation is taking the role of the true transport
Even in 3G networks Look for latest ITU drafts for Ethernet aggregation in 3G networks